proxmox/docs/03-deployment/MIGRATE_503GB_R630_TO_128_256GB_SERVERS.md

Migrate workload off 503 GB R630 → r630-05 through r630-08 (256 GB each)

Last Updated: 2026-01-31
Document Version: 1.0
Status: Active Documentation

---

Last updated: 2026-01-31
Goal: Move all workload off the 503 GB R630 (r630-01) onto r630-05, r630-06, r630-07, r630-08, each with 256 GB RAM, for the HA cluster.

---

Current state (reference)

Host	IP	RAM (from health check)	LXC count
ml110	192.168.11.10	125 GB	17
r630-01	192.168.11.11	503 GB	69
r630-02	192.168.11.12	251 GB	10

The 503 GB server is r630-01 (192.168.11.11). That is the source host to migrate workload from.

---

Target layout: r630-05 through r630-08 (256 GB each)

Host	IP (planned)	RAM	Role / use
r630-05	192.168.11.15	256 GB	HA compute node 1
r630-06	192.168.11.16	256 GB	HA compute node 2
r630-07	192.168.11.17	256 GB	HA compute node 3
r630-08	192.168.11.18	256 GB	HA compute node 4

• 4 nodes × 256 GB meets the HA cluster target (3–4 nodes, 128–256 GB per node).
• Assign IPs 192.168.11.15–18 (or your chosen range) when the hosts are racked and on the same VLAN as the rest of the cluster.
• Migrate workload from r630-01 (and optionally from ml110/r630-02) onto these four nodes.

---

Target state

• No single server with 503 GB holding all workload.
• r630-05, r630-06, r630-07, r630-08 as the primary HA compute pool, each 256 GB RAM.
• Workload spread across the four nodes; critical services on nodes that participate in HA.

---

Phase 1: Inventory and plan

1. List everything on the 503 GB host (r630-01).

   • From project root:
     ./scripts/quick-proxmox-inventory.sh
     or SSH:
     ssh root@192.168.11.11 "pct list; qm list"
   • Note: VMID, name, RAM/CPU, and whether it’s critical (e.g. NPMplus 10233, RPC, Blockscout, DBIS, etc.).

2. Decide destination per VM/container.

   • ml110 (125 GB): optional for lighter containers.
   • r630-02 (251 GB): optional overflow; long-term can also be migrated to r630-05..08.
   • r630-05, r630-06, r630-07, r630-08 (256 GB each): primary targets; spread workload from r630-01 across all four.

3. Strategy.

   • Add r630-05, r630-06, r630-07, r630-08 to the cluster (256 GB each, IPs e.g. 192.168.11.15–18). Migrate workload from r630-01 to these four nodes.
   • When r630-01 is empty: power off, reduce RAM (remove DIMMs) if reusing; otherwise decommission.

---

Phase 2: Migrate workload off r630-01

1. Storage.
   Today there is no shared storage; each VM/container’s disk lives on the host. So migration is:

   • LXC: pct migrate <vmid> <target-node> (or stop → backup → restore on target).
   • QEMU: qm migrate <vmid> <target-node> (live if storage allows) or stop → backup → restore on target.

2. Order (suggested).

   • Migrate non-critical containers first (test, dev, duplicate roles).
   • Then critical ones: NPMplus (10233), RPC-related, Blockscout, DBIS, etc. Do these in a maintenance window if you want minimal impact.

3. Example – migrate one LXC to r630-05 (or r630-06, r630-07, r630-08).
   From any node with cluster access, or from r630-01:

   # List containers on r630-01
   ssh root@192.168.11.11 "pct list"
   
   # Migrate LXC 10234 to r630-05 (target must have storage; use r630-05..08 as needed)
   pct migrate 10234 r630-05 --restart
   

   If pct migrate fails (e.g. no shared storage), use backup on source → restore on target:

   # On r630-01: backup
   pct backup <vmid> backup-<vmid>.tar.gz --compress zstd
   
   # Copy to r630-05 (or shared storage), then on r630-05:
   pct restore <vmid> backup-<vmid>.tar.gz
   # Reconfigure network (IP, etc.) if needed, then start.
   

4. After each move:
   Check service on the new host (IP, DNS, NPMplus proxy targets, etc.).

---

Phase 3: Downsize r630-01 to 128–256 GB

1. When r630-01 has no (or minimal) workload:
   Power off r630-01 (or put in maintenance).

2. Reseat / remove DIMMs so total RAM is 128 GB or 256 GB (per your choice).

   • Use Dell docs / R630 Owner’s Manual for population rules (which slots to leave populated for 128 or 256 GB).
   • Keep DIMMs you pull for use in other R630s (to bring them to 128–256 GB).

3. Power on r630-01, confirm RAM in BIOS and in Proxmox (e.g. free -h).

4. Repeat for r630-02 if it currently has 251 GB and you want it at 128–256 GB; use DIMMs from r630-01 if needed.

---

Phase 4: Balance and HA readiness

• Ensure no single node is overloaded (CPU/RAM).
• Document final RAM per server: e.g. ml110 125 GB, r630-01 256 GB, r630-02 256 GB, (optional) r630-03 256 GB.
• When you introduce shared storage (Ceph or NFS) and Proxmox HA, these 128–256 GB nodes will match the “128–256 GB per server” HA target.

---

Quick reference

Step	Action
1	Inventory r630-01: ssh root@192.168.11.11 "pct list; qm list" or ./scripts/quick-proxmox-inventory.sh
2	Choose destinations: r630-05, r630-06, r630-07, r630-08 (256 GB each); ml110/r630-02 optional.
3	Migrate LXC: pct migrate <vmid> <target-node> or backup/restore.
4	Migrate QEMU: qm migrate <vmid> <target-node> or backup/restore.
5	When r630-01 is empty: power off, reduce RAM to 128–256 GB, power on.
6	Add r630-05..08 to cluster (256 GB each); optionally downsize r630-02 using DIMMs from r630-01.

Target: All workload off the 503 GB R630 onto r630-05, r630-06, r630-07, r630-08 (256 GB each) for the HA cluster.